The 23rd Conference on Weather Analysis and Forecasting (WAF)/1

Satellite measurements in various spectral channels are assimilated through the JCSDA's Community Radiative Transfer Model (CRTM) on the NCEP Gridpoint Statistical Interpolation (GSI) (Wu et al. 2002). Currently, satellite measurements over ocean have been successfully utilized to improve numerical weather prediction (NWP). However, it is found that the amount of satellite data assimilated over land in the GSI/CRTM is far less than over ocean. One of the chief reasons is that there is a much larger cold bias in the current NCEP operational Global Forecast System (GFS) predicted land surface skin temperature (LST) over desert and arid regions during daytime in the warm season. LST predicted by the GFS is a critical factor to determine brightness temperature (Tb) simulation for satellite surface sensitive channels. After having compared LST in GFS/GDAS with that from GOES-derived or SURFRAD observations in summer, we found GFS/GDAS has a substantial cold bias of more than -12K over the arid western CONUS during daytime (Zheng et al., 2008). With such a large cold bias in LST, the CRTM simulates unreasonable brightness temperatures, thus most of satellite data are rejected in the GSI/CRTM analysis step, especially for surface sensitive satellite channels. An investigation in GFS testing has revealed a major cause of the cold daytime LST bias is in the treatment for roughness lengths, particularly thermal roughness length (z0t) in the physics of surface turbulent heat transfer. In this study, alternative formulations of momentum and thermal roughness lengths developed by Zeng et al. (personal communication) are tested to reduce the GFS warm season mid-day cold bias in LST. The impact of new roughness changes on the brightness temperature calculation in GSI is investigated to improve satellite data assimilation.